Dissertation (Ph.D.) -- University of Texas Southwestern Medical Center at Dallas, 2006. / Vita. Bibliography: pp. 140-153

Rotator cuff disorders encompass a range of impairments from tendinopathy to partialor full-thickness rotator cuff tears, and represent the largest subgroup of shoulder pain. Rotator cuff tears, most of which are atraumatic, are common in adults with shoulder pain and are strongly associated with increasing age. Conservative treatment including physiotherapy is the first-line treatment, but some patients do not respond, and ultimately require surgery. Early predictions of response could allow individuals’ care pathways to be optimised, preventing unnecessary delays and suffering and benefiting patients and healthcare providers alike. My primary aim was to develop a prognostic model for the outcome of physiotherapy in adults with painful atraumatic partial-thickness tears (PTTs) of the rotator cuff. This was addressed by a prospective prognostic model study. The study was underpinned by a systematic review of prognostic models in adults undergoing physiotherapy for painful rotator cuff disorders and was further informed and complemented by the following work: the development and validation of the physiotherapy protocol for the prognostic study; the identification, selection and definition of the candidate prognostic factors for the prognostic study; the estimation of the Minimal Important Difference (MID) of the study’s primary outcome measure (the Western Ontario Rotator Cuff Index, WORC); and an exploratory responder analysis of the WORC outcome scores. The prognostic systematic review, prognostic study, MID analysis and responder analysis are original contributions to knowledge. The prognostic systematic review revealed important methodological deficiencies in the five included studies, and no clinically usable model. No study addressed a distinct PTT population. The process of identifying factors for my own prognostic model study revealed a lack of knowledge about the prognostic relevance of factors. All of the candidate models I explored in my prognostic study (n sample = 65, n analysed = 61) had low performance and precision. The estimated MID of the WORC was -300. The responder analysis resulted in different proportions of responders to treatment depending on the responder definition. My results highlight the difficulties involved in predicting outcomes in the field of shoulder pain and rotator cuff disorders, and the need for methodologically sound prognosis research.

Lung cancer is the leading cause of cancer related death worldwide; despite recent treatment developments survival rates remain poor and are closely related to the patient’s clinical stage. Even among patients with early-stage lung cancer, which is amenable to surgical resection, prognosis is highly variable; some go on to live disease-free for many years whereas others quickly recur. Although post-operative chemotherapy is available it has associated morbidities and it is unclear which patients would benefit; therefore, there is a need for more effective stratification of patients. The adenocarcinoma sub-type of lung cancer is known to be morphologically heterogeneous however the majority of observed growth patterns, assessed by light microscopy, can be characterised into one of five formations: lepidic, papillary, acinar, solid and micropapillary. The morphology of each tumour has been proposed as a marker of prognosis and several studies have published a link between the most prevalent growth pattern and prognosis; suggesting those with predominantly solid or micropapillary tumours to have the least favourable outcomes. Indeed, it is now recommended that the proportion of each growth pattern and the predominant growth pattern should be reported for all resected lung adenocarcinomas; although no differential treatments have been recommended based on this assessment. The aim of this study was to determine whether combining the analysis of clinicopathological; morphological; and candidate protein, molecular genetic and transcriptomic characteristics in a single cohort of 208 early-stage, resected, adenocarcinomas with clinical follow-up could be used to identify a subset of patients at high risk of recurrence. Comprehensive morphological analysis was carried out including the presence, proportion and number of individual growth patterns; the predominant growth pattern as well as features previously associated with tumour grade (the presence of large numbers of mitotic figures, apoptotic bodies, inflammatory cells, prominent nucleoli, pleomorphic tumour cells, dyscohesive tumour cells and large amounts of necrosis and scar tissue within the tumour). In addition, gene expression was assessed using a panel of 31 cell-cycle related genes, EGFR and KRAS mutation status was determined, and EGFR and TTF1 protein expression investigated. In this study the predominant growth pattern defined by histopathology showed no ability to identify a group of patients with a poorer prognosis either in univariable or multivariable analysis. Univariable analysis identified nodal status [hazard ratio of N1 compared to N0 was 2.16 (95% CI 1.48 to 3.16, p< 0.0005)], clinical stage [hazard ratios of stage IIa and IIb compared to stage Ia were 3.15 (95% CI 1.73 to 5.73, p< 0.0005) and 2.22 (95% CI 1.10 to 4.48, p= 0.025) respectively], the presence of a significant amount of the papillary growth pattern [the hazard ratio of those with less than 8.5% papillary pattern was 0.657 (95% CI 0.44 to 0.98, p= 0.035)], and overall tumour grade score (including an assessment of necrosis, mitosis, apoptosis, nucleoli, scar tissue and inflammatory cells) [hazard ratio 1.71 (95% CI 1.14 to 2.56, p= 0.008)] as significantly associated with prognosis. Multivariable analysis using Cox’s proportional hazards model identified clinical stage (p< 0.0005), the presence of a significant amount of the papillary growth pattern (p= 0.048) and the presence of large numbers of mitotic figures (p=0.029) and apoptotic bodies (p= 0.015) as independently associated with disease specific survival; although after correction for type I errors only clinical stage remained significantly associated with prognosis with patients with stage Ia disease having a significantly better outcomes [hazard ratio 0.418 (95% CI 0.20 to 0.86)]. Classification and regression tree analysis (CART) was used to further explore the data and to develop decision trees for the prognostication of early-stage lung adenocarcinoma patients. Receiver operating characteristic analysis based on 5- year survival showed a minimal improvement in the area under the curve between a model utilizing currently available clinicopathologic characteristics only [nodal status and lesion size, (area under the curve 0.704, 95% CI 0.631 to 0.777)] and one including growth pattern characteristics [area under the curve 0.725, 95% CI 0.654 to 0.796]. The greatest improvement in prognostic accuracy was observed when gene expression analysis was included in the analysis [area under the curve 0.749, 95% CI 0.673 to 0.825]; however even this showed very little impact compared to routinely used clinicopathologic variables. This analysis suggests that the recommended characterisation of lung adenocarcinoma histology is not a robust predictor of patient outcomes; even a broader model which also included indicators of tumour grade and molecular characteristics was unable to identify a model sufficiently robust to implement into clinical practice and thereby potentially alter patient treatment. Currently routinely collected clinical characteristics; including nodal status, size and clinical stage; continue to provide the most robust method of prognostication and detailed and time-consuming morphological analysis offers no significant benefit to the patient.

Abstract In tumourigenesis, genetic alterations accumulate in the genes responsible for cell growth, proliferation and DNA repair: proto-oncogenes, tumour suppressor and DNA repair genes. Inactivation of tumour suppressor gene function is commonly recognised as a deletion of one of the two alleles; LOH, loss of heterozygosity. In the present study, LOH of several chromosomal regions was studied in both breast and ovarian cancer. LOH for chromosome region 11q was examined in a large breast cancer consortium cohort (N = 988) and in a Finnish ovarian cancer cohort (N = 78), and the clinical significance of these alterations was evaluated. In breast cancer, LOH of the studied markers at 11q23.1, harbouring approximately 2 Mb of DNA, was seen to be associated with shortened cancer-specific survival. Two candidate genes, ATM (the ataxia telangiectasia disorder gene) and DDX10 (a putative RNA helicase gene) map to this chromosomal region. In ovarian cancer, LOH at 11q23.1–q24 was assigned mainly to two chromosomal regions, A and B, which are proximal and distal to 11q23.2–q23.3, respectively. Only the distal B region was seen to be associated with an aggressive disease course. Therefore, it appears that inactivation of the ATM or DDX10 genes is not crucial for determining the outcome of ovarian cancer. The CHK1 gene at 11q24, encoding a protein kinase required for DNA damage checkpoint function, is a putative target gene at the B region. On the basis of the present results, the main TSG in the studied region involved in the progression of breast cancer maps to 11q23.1 and the corresponding gene for ovarian cancer more distally to 11q23.3-q24. In addition, LOH at 3p, 6q, 8p, 11p, 16q and 17p was examined and their role in the genetic evolution of ovarian cancer was evaluated. Of the studied chromosomal regions, LOH at 17p appeared to be an early event and LOH at 16q24.3, 11q23.3/q24 and 11p appear to occur later in the progression of ovarian cancer.

The theme of this dissertation is to construct heterogeneous biomarker networks using graphical models for understanding disease progression and prognosis. Biomarkers may organize into networks of connected regions. Substantial heterogeneity in networks between individuals and subgroups of individuals is observed. The strengths of network connections may vary across subjects depending on subject-specific covariates (e.g., genetic variants, age). In addition, the connectivities between biomarkers, as subject-specific network features, have been found to predict disease clinical outcomes. Thus, it is important to accurately identify biomarker network structure and estimate the strength of connections. Graphical models have been extensively used to construct complex networks. However, the estimated networks are at the population level, not accounting for subjects’ covariates. More flexible covariate-dependent graphical models are needed to capture the heterogeneity in subjects and further create new network features to improve prediction of disease clinical outcomes and stratify subjects into clinically meaningful groups. A large number of parameters are required in covariate-dependent graphical models. Regularization needs to be imposed to handle the high-dimensional parameter space. Furthermore, personalized clinical symptom networks can be constructed to investigate co-occurrence of clinical symptoms. When there are multiple biomarker modalities, the estimation of a target biomarker network can be improved by incorporating prior network information from the external modality. This dissertation contains four parts to achieve these goals: (1) An efficient l0-norm feature selection method based on augmented and penalized minimization to tackle the high-dimensional parameter space involved in covariate-dependent graphical models; (2) A two-stage approach to identify disease-associated biomarker network features; (3) An application to construct personalized symptom networks; (4) A node-wise biomarker graphical model to leverage the shared mechanism between multi-modality data when external modality data is available. In the first part of the dissertation, we propose a two-stage procedure to regularize l0-norm as close as possible and solve it by a highly efficient and simple computational algorithm. Advances in high-throughput technologies in genomics and imaging yield unprecedentedly large numbers of prognostic biomarkers. To accommodate the scale of biomarkers and study their association with disease outcomes, penalized regression is often used to identify important biomarkers. The ideal variable selection procedure would search for the best subset of predictors, which is equivalent to imposing an l0-penalty on the regression coefficients. Since this optimization is a non-deterministic polynomial-time hard (NP-hard) problem that does not scale with number of biomarkers, alternative methods mostly place smooth penalties on the regression parameters, which lead to computationally feasible optimization problems. However, empirical studies and theoretical analyses show that convex approximation of l0-norm (e.g., l1) does not outperform their l0 counterpart. The progress for l0-norm feature selection is relatively slower, where the main methods are greedy algorithms such as stepwise regression or orthogonal matching pursuit. Penalized regression based on regularizing l0-norm remains much less explored in the literature. In this work, inspired by the recently popular augmenting and data splitting algorithms including alternating direction method of multipliers, we propose a two-stage procedure for l0-penalty variable selection, referred to as augmented penalized minimization-L0 (APM-L0). APM-L0 targets l0-norm as closely as possible while keeping computation tractable, efficient, and simple, which is achieved by iterating between a convex regularized regression and a simple hard-thresholding estimation. The procedure can be viewed as arising from regularized optimization with truncated l1 norm. Thus, we propose to treat regularization parameter and thresholding parameter as tuning parameters and select based on cross-validation. A one-step coordinate descent algorithm is used in the first stage to significantly improve computational efficiency. Through extensive simulation studies and real data application, we demonstrate superior performance of the proposed method in terms of selection accuracy and computational speed as compared to existing methods. The proposed APM-L0 procedure is implemented in the R-package APML0. In the second part of the dissertation, we develop a two-stage method to estimate biomarker networks that account for heterogeneity among subjects and evaluate the network’s association with disease clinical outcome. In the first stage, we propose a conditional Gaussian graphical model with mean and precision matrix depending on covariates to obtain subject- or subgroup-specific networks. In the second stage, we evaluate the clinical utility of network measures (connection strengths) estimated from the first stage. The second stage analysis provides the relative predictive power of between-region network measures on clinical impairment in the context of regional biomarkers and existing disease risk factors. We assess the performance of the proposed method by extensive simulation studies and application to a Huntington’s disease (HD) study to investigate the effect of HD causal gene on the rate of change in motor symptom through affecting brain subcortical and cortical grey matter atrophy connections. We show that cortical network connections and subcortical volumes, but not subcortical connections are identified to be predictive of clinical motor function deterioration. We validate these findings in an independent HD study. Lastly, highly similar patterns seen in the grey matter connections and a previous white matter connectivity study suggest a shared biological mechanism for HD and support the hypothesis that white matter loss is a direct result of neuronal loss as opposed to the loss of myelin or dysmyelination. In the third part of the dissertation, we apply the methodology to construct heterogeneous cross-sectional symptom networks. The co-occurrence of symptoms may result from the direct interactions between these symptoms and the symptoms can be treated as a system. In addition, subject-specific risk factors (e.g., genetic variants, age) can also exert external influence on the system. In this work, we develop a covariate-dependent conditional Gaussian graphical model to obtain personalized symptom networks. The strengths of network connections are modeled as a function of covariates to capture the heterogeneity among individuals and subgroups of individuals. We assess the performance of the proposed method by simulation studies and an application to a Huntington’s disease study to investigate the networks of symptoms in different domains (motor, cognitive, psychiatric) and identify the important brain imaging biomarkers associated with the connections. We show that the symptoms in the same domain interact more often with each other than across domains. We validate the findings using subjects’ measurements from follow-up visits. In the fourth part of the dissertation, we propose an integrative learning approach to improve the estimation of subject-specific networks of target modality when external modality data is available. The biomarker networks measured by different modalities of data (e.g., structural magnetic resonance imaging (sMRI), diffusion tensor imaging (DTI)) may share the same true underlying biological mechanism. In this work, we propose a node-wise biomarker graphical model to leverage the shared mechanism between multi-modality data to provide a more reliable estimation of the target modality network and account for the heterogeneity in networks due to differences between subjects and networks of external modality. Latent variables are introduced to represent the shared unobserved biological network and the information from the external modality is incorporated to model the distribution of the underlying biological network. An approximation approach is used to calculate the posterior expectations of latent variables to reduce time. The performance of the proposed method is demonstrated by extensive simulation studies and an application to construct gray matter brain atrophy network of Huntington’s disease by using sMRI data and DTI data. The estimated network measures are shown to be meaningful for predicting follow-up clinical outcomes in terms of patient stratification and prediction. Lastly, we conclude the dissertation with comments on limitations and extensions.

Thesis advisor: Samuel Richardson / HbA1c levels are the most frequently used test for diagnosis and prognosis of diabetes mellitus. Recent studies have shown the biases this test has in particular cohorts, that was not noted when it was originally accepted by the American Diabetes Association in 2008. This study examined how these biases affect HbA1c’s ability as a predictor for complications that arise due to diabetes in specific cohorts, those of ethnicity, age, weight, and other patient attributes, compared to other established diabetes prognosis tests. We discovered that both glucose and HbA1c share similar biases as predictors for particular cohorts (the high glucose, high BMI, Asian, African, and Hispanic descent cohorts), HbA1c works better as a predictor when it is combined with the results of a glucose test and more characteristics of the patient compared to a HbA1c test alone with fewer variables, and glucose and HbA1c are better predictors for different diseases, respectively, that may arise due to diabetes mellitus. / Thesis (BA) — Boston College, 2020. / Submitted to: Boston College. College of Arts and Sciences. / Discipline: Departmental Honors. / Discipline: Economics.

Background: Anemia is common in congestive heart failure, and it has been associated with poor prognosis. The effect of anemia on functional ability in heart failure has not been described. We evaluated the relationship of anemia, physical disability, and survival in patients with heart failure. Methods and Results: One-year longitudinal study of 567 non-disabled, hospitalized heart failure patients, age ≥65 years, enrolled in the Italian Group of Pharmacoepidemiology in the Elderly Study. Anemia was defined according to the World Health Organization criteria. Physical disability was defined as dependence in performing at least 2 basic activities of daily living. After adjustment for disease severity and health-related variables, anemia was associated with higher risk of disability (odds ratio = 2.17; 95% confidence interval [CI] = 1.12-4.24). After stratification according to gender, a strong relationship of anemia and risk of disability persisted in women, but it was reduced in men. Anemic women were significantly more likely to die during the follow-up, even after adjustment for potential confounders (hazard ratio = 2.33; CI = 1.02-5.30). Conclusion: Anemia is a predictor of physical disability in older heart failure patients, and in women anemia is associated with increased mortality.

Cardiac arrest is a common cause of coma with frequent poor outcomes. Palliative medicine teams are often called upon to discuss the scope of treatment and future care in cases of anoxic brain injury. Understanding prognostic tools in this setting would help medical teams communicate more effectively with patients’ families and caregivers and may promote improved quality of life overall. This article reviews multiple tools that are useful in determining outcomes in the setting of postarrest anoxic brain injury.

Bearings are critical in today’s mechanisms, and their reliability is continuously improving. Yet, working under high loads for long periods, bearings will degrade and eventually fail. An unpredicted bearing failure can lead to total and catastrophic failures of machines and may even lead to human injuries that result in substantial economic losses and reductions in production. Determining a bearing’s remaining useful life (RUL) has become an important topic in many industrial fields. Vibration signals are the most used representation for understanding a bearing’s health status. Using different algorithms, time-domain vibration signals can be transformed into time-frequency domain signals that help indicate a bearing’s status. For instance, this thesis investigates spectrograms and scalograms to visually represent a bearing’s health condition using a short-time Fourier transform (STFT) and a continuous wavelet transform (CWT). Both representations are plotted as a function of time and frequency and can detect the bearing’s working condition. However, spectrograms are advantageous in revealing frequency changes along the time axis, while scalograms facilitate the detection of abrupt changes. Combined with a convolutional neural network (CNN), these plots can be used to interpret bearing RUL. The strength of CNNs lie in their ability to identify and detect features in images, including such tasks as image classification, using share-weight architectures, convolutional layers, and kernels. This thesis explores CNNs combined with spectrograms and scalograms using the PRONOSTIA dataset to perform bearing RUL predictions and explore relationships between prognosis and diagnosis for bearing faults analysis.

Improved survival in patients with hematological malignancy (HM) admitted to the intensive care unit (ICU) has largely been reported in uncontrolled cohorts from single academic institutions. We compared hospital mortality between 147 patients with HM and 147 general medical admissions to five non-specialist ICUs. The proportion of patients surviving to hospital discharge was significantly worse in patients with HM (27% vs. 56%; p < 0.001). Six-month and 1-year survival in patients with HM was 21% and 18%, respectively. HM, greater age, mechanical ventilation (MV) and acute physiology and chronic health evaluation (APACHE) II score were independent predictors of poor outcome. For patients with HM, culture proven infection, age, MV and inotropes were negative predictors. Disease-specific factors including hematological diagnosis, neutropenia, remission status, prior stem cell transplant, time from diagnosis to admission and degree of prior treatment were not predictive. Overall survival of patients with HM was worse than that recently reported from specialist units.